Interview Prep
Interview questions, roadmaps and practice sets
Tutorials
Free tutorials on 30+ topics
Blog
Career guides, skills and interview tips
Free PDFs
Notes, cheat sheets and question banks
Certificates
Free skill tests and course certificates
Tools & Whiteboard
Free tools you can use right in the browser
Practice Tests
Mock tests, MCQs and quick Q&A
Free LMS
Learn on the Uncodemy LMS
Choose the best training program according to your career.
Improve the performance of your business with our innovative Industrial Training Program.
Read More →
Our instructor-led training classes are designed to make you an expert.
Read More →
Real-life case studies are a great way to demonstrate your practical knowledge.
Read More →
Explore blogs, latest technologies and industry updates with practical examples.
Read More →Browse all available training programs and choose the right one for your career.
Read More →Know more about who we are and what we stand for.
Discover our mission to equip individuals with essential IT skills and industry expertise.
Read More →
Free tutorials, interview questions, tools and calculators to help you learn.
Explore →Learn, practice and plan your career — bilkul free.
Book a free video counselling session and pick the right course for you.
Read More →GPT-4 Turbo (late 2023) and GPT-4o (May 2024) are two steps in OpenAI's GPT-4 line. This page sets out what changed between them, using OpenAI's own published figures with their dates, and shows how to estimate the cost of a workload on each.


| Model | Announced | Context window | Notes |
|---|---|---|---|
| GPT-4 | 14 March 2023 | 8K tokens (32K variant) | Text output; image input added later |
GPT-4 Turbo (preview, gpt-4-1106-preview) | 6 November 2023, OpenAI DevDay | 128K | Lower price than GPT-4, JSON mode |
GPT-4 Turbo with Vision (gpt-4-turbo-2024-04-09) | April 2024 (general availability) | 128K | Image input in the same model; knowledge cutoff December 2023 |
GPT-4o (gpt-4o-2024-05-13) | 13 May 2024 | 128K | "o" for omni: text, image and audio in one model |
GPT-4o (gpt-4o-2024-08-06) | August 2024 | 128K | Lower price, Structured Outputs, up to 16,384 output tokens |
Sources: OpenAI's DevDay announcement (November 2023), "Hello GPT-4o" (May 2024) and the OpenAI model documentation.
Before GPT-4o, ChatGPT's voice mode chained three models: speech-to-text, GPT-4 for the reply, and text-to-speech. OpenAI's May 2024 post gave average voice-mode latencies of 2.8 seconds with GPT-3.5 and 5.4 seconds with GPT-4. GPT-4o was trained end to end across text, vision and audio, and OpenAI reported it can respond to audio in as little as 232 ms, about 320 ms on average. GPT-4 Turbo accepts images but has no native audio.
OpenAI's launch claim was that GPT-4o in the API is "2x faster, half the price, and has 5x higher rate limits" than GPT-4 Turbo. Real throughput depends on load, region, prompt length and output length, so measure it on your own requests before relying on it.
GPT-4 and GPT-4 Turbo use the cl100k_base tokenizer; GPT-4o uses o200k_base. For many non-English languages the same text becomes fewer tokens, which means lower cost and more room in the context window. In its GPT-4o announcement OpenAI's examples included Hindi using 2.9 times fewer tokens and Tamil 3.3 times fewer. English improves only slightly.
OpenAI published these text evaluation scores at the GPT-4o launch (May 2024, using its simple-evals setup):
| Benchmark | What it tests | GPT-4o | GPT-4 Turbo (2024-04-09) |
|---|---|---|---|
| MMLU | General knowledge across 57 subjects | 88.7 | 86.5 |
| GPQA | Graduate-level science questions | 53.6 | 48.0 |
| MATH | Competition maths problems | 76.6 | 72.6 |
| HumanEval | Writing Python functions | 90.2 | 87.1 |
Two cautions. These are the vendor's own numbers, and published benchmarks can be partly contaminated by training data. A gap of two or three points rarely decides which model is better for your task.
| Model | Input (USD per 1M tokens) | Output (USD per 1M tokens) |
|---|---|---|
| GPT-4 (8K) | 30.00 | 60.00 |
| GPT-4 Turbo | 10.00 | 30.00 |
| GPT-4o (May 2024) | 5.00 | 15.00 |
GPT-4o (2024-08-06) | 2.50 | 10.00 |
Prices are from OpenAI's announcements and pricing page at each release. They change, and discounts such as batch processing and cached input apply, so check the current pricing page before budgeting.
Suppose a support assistant handles 10,000 requests, each with about 1,500 input tokens (instructions plus the customer's history) and 500 output tokens.
PRICES = { # USD per 1M tokens (input, output), launch prices
"gpt-4 (8k)": (30.00, 60.00),
"gpt-4-turbo": (10.00, 30.00),
"gpt-4o (May 2024)": (5.00, 15.00),
"gpt-4o-2024-08-06": (2.50, 10.00),
}
requests, tin, tout = 10_000, 1_500, 500
for model, (pin, pout) in PRICES.items():
cost = requests * (tin * pin + tout * pout) / 1_000_000
print(f"{model:18s} ${cost:>7,.2f}")
Output:
gpt-4 (8k) $ 750.00
gpt-4-turbo $ 300.00
gpt-4o (May 2024) $ 150.00
gpt-4o-2024-08-06 $ 87.50
Output tokens cost two to four times as much as input tokens, so limiting response length usually saves more than trimming the prompt. For non-English traffic, count tokens with the right tokenizer (the tiktoken library supports both) because GPT-4o will often need fewer.
If an existing app was built on GPT-4 Turbo, re-run your tests after switching. Output style and formatting can differ even when quality is similar, and prompts that relied on one model's habits may need adjustment.
Both have been followed by newer releases: GPT-4o mini (July 2024), GPT-4.1 (April 2025) and the GPT-5 family (from August 2025). In ChatGPT, GPT-4 was retired at the end of April 2025 and replaced by GPT-4o. The comparison above is still useful for understanding the shift to native multimodal models and lower prices, and for maintaining apps pinned to these model versions. For the wider field see GPT-4, Claude 3 and Gemini compared and newer LLMs to know. If you are new to the topic, start with an introduction to LLMs; for building with these APIs, see LLM application development.
On OpenAI's own published text benchmarks from May 2024, GPT-4o scored a little higher than GPT-4 Turbo, and it was priced lower and described as faster. It also handles audio and images natively. For a specific task, test both on your own examples.
Omni. OpenAI used the name to signal that one model handles text, audio and images as input and output.
No. OpenAI has released newer models since, including GPT-4.1 in April 2025 and the GPT-5 family from August 2025. Check OpenAI's models and deprecations pages for what is currently available.